import os
# Change dir
os.chdir('C:\\Users\\ThalesFerreira\\projetos python\\data_view\\shipping_data')
#import models and libs
import joblib

# treat data
import pandas as pd 
import numpy as np
# data scaling data
from sklearn.preprocessing import StandardScaler
# data labelencoder
from sklearn.preprocessing import LabelEncoder
# data onehotencoder
from sklearn.preprocessing import OneHotEncoder
# multilinear correlations
from statsmodels.stats.outliers_influence import variance_inflation_factor
# anomalies
from sklearn.neighbors import LocalOutlierFactor

# plot graphcs of data
import matplotlib.pyplot as plt
import seaborn as sns

# calcular score do modelo
from sklearn.metrics import accuracy_score, classification_report
#Metricas para avaliação do modelo
from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score, confusion_matrix
#Divisão de bases e crossvalidation.
from sklearn.model_selection import train_test_split, cross_val_score, KFold
# Confusion matrix
from yellowbrick.classifier import ConfusionMatrix

#Anova
from scipy.stats import f_oneway
#normalization teste
from scipy.stats import shapiro
#Turkey teste
from statsmodels.stats.multicomp import MultiComparison

#ramdom parameters
from sklearn.model_selection import RandomizedSearchCV
# tests models
from sklearn.model_selection import GridSearchCV
# Logistic Regression models
from sklearn.linear_model import LogisticRegression
#Bayes
from sklearn.naive_bayes import GaussianNB
# Decision Tree
from sklearn.tree import DecisionTreeClassifier
# KNN
from sklearn.neighbors import KNeighborsClassifier
# Random Forest
from sklearn.ensemble import RandomForestClassifier
# Neural Network
from sklearn.neural_network import MLPClassifier
# SVM
from sklearn.svm import SVC

df = pd.read_csv('./base/Train.csv',sep=',',decimal='.',encoding='utf-8')

df.shape

(10999, 12)

df.describe(include='all').transpose()

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10999 entries, 0 to 10998
Data columns (total 12 columns):
 #   Column               Non-Null Count  Dtype 
---  ------               --------------  ----- 
 0   ID                   10999 non-null  int64 
 1   Warehouse_block      10999 non-null  object
 2   Mode_of_Shipment     10999 non-null  object
 3   Customer_care_calls  10999 non-null  int64 
 4   Customer_rating      10999 non-null  int64 
 5   Cost_of_the_Product  10999 non-null  int64 
 6   Prior_purchases      10999 non-null  int64 
 7   Product_importance   10999 non-null  object
 8   Gender               10999 non-null  object
 9   Discount_offered     10999 non-null  int64 
 10  Weight_in_gms        10999 non-null  int64 
 11  Reached.on.Time_Y.N  10999 non-null  int64 
dtypes: int64(8), object(4)
memory usage: 1.0+ MB

df.isna().sum()

ID                     0
Warehouse_block        0
Mode_of_Shipment       0
Customer_care_calls    0
Customer_rating        0
Cost_of_the_Product    0
Prior_purchases        0
Product_importance     0
Gender                 0
Discount_offered       0
Weight_in_gms          0
Reached.on.Time_Y.N    0
dtype: int64

df.nunique()

ID                     10999
Warehouse_block            5
Mode_of_Shipment           3
Customer_care_calls        6
Customer_rating            5
Cost_of_the_Product      215
Prior_purchases            8
Product_importance         3
Gender                     2
Discount_offered          65
Weight_in_gms           4034
Reached.on.Time_Y.N        2
dtype: int64

df.select_dtypes(include=object).nunique()

Warehouse_block       5
Mode_of_Shipment      3
Product_importance    3
Gender                2
dtype: int64

df.sample(5)

df['Reached.on.Time_Y.N'].value_counts()

Reached.on.Time_Y.N
1    6563
0    4436
Name: count, dtype: int64

# Delete colum ID
if 'ID' in df.columns:
    df.drop(columns=['ID'],inplace=True)

#Rename columns
if 'Reached.on.Time_Y.N' in df.columns:
    df.rename(columns={'Reached.on.Time_Y.N':'on_time'},inplace=True)

plt.Figure(figsize=(7,7))
plot = sns.countplot(data=df,x='on_time',hue='on_time')
plot.axhline(y=df.shape[0]/2,color='red')
plt.show()

df.select_dtypes(exclude=object).hist(layout=(3,3), figsize=(20,7))
plt.show()

df.select_dtypes(exclude=object).plot(kind='box', subplots=True, layout=(2,4), figsize=(20,7))
plt.show()

#reaching At the time
plt.figure(figsize=(12, 6))
for i, col in enumerate(df.drop('on_time',axis=1).select_dtypes(exclude=object).columns):
    plt.subplot(2, 3, i+1)
    sns.boxplot(data=df, x='on_time', y=col)
plt.suptitle("Distribution per arrived on time")
plt.tight_layout()
plt.show()

#reaching At the time count category
plt.figure(figsize=(12, 8))
for i, col in enumerate(df.drop('on_time',axis=1).select_dtypes(include=object).columns):
    plt.subplot(2, 2, i+1)
    sns.countplot(data=df,x=col,hue='on_time')
    #sns.barplot(data=df, x='on_time', y=col,estimator=np.count_nonzero)
plt.suptitle("Distribution per arrived on time")
plt.tight_layout()
plt.show()

#reaching At the time mean category
plt.figure(figsize=(12, 8))
for i, col in enumerate(df.drop('on_time',axis=1).select_dtypes(include=object).columns):
    plt.subplot(2, 2, i+1)
    sns.barplot(data=df, x='on_time', y=col,estimator=np.mean)
plt.suptitle("Distribution per arrived on time")
plt.tight_layout()
plt.show()

lof = LocalOutlierFactor(n_neighbors=20,contamination=0.05)

lof_ind = np.where(lof.fit_predict(df.select_dtypes(exclude=object))==-1)

df['anom'] = 0
df.loc[lof_ind[0],'anom'] = 1

plt.figure(figsize=(8,8))
plt.subplot(2, 2, 1)
sns.scatterplot(data=df,y='Discount_offered',x='Cost_of_the_Product',hue='anom')
plt.subplot(2, 2, 2)
sns.scatterplot(data=df,y='Prior_purchases',x='Cost_of_the_Product',hue='anom')
plt.subplot(2, 2, (3,4))
sns.scatterplot(data=df,y='Prior_purchases',x='Discount_offered',hue='anom')
plt.show()

df.drop(columns=['anom'],inplace=True)

for col in df.select_dtypes(include=object).columns:
    df[col] = LabelEncoder().fit_transform(df[col])

mask = np.triu(np.ones_like(df.select_dtypes(exclude=object).corr(), dtype=np.bool_))
plt.figure(figsize=(15,15))
sns.heatmap(df.select_dtypes(exclude=object).corr(method='spearman')
            ,mask=mask,fmt='.2f' ,square = True
            ,annot=True, vmin=-1, vmax=1)
plt.show()

plt.figure(figsize=(15,15))
sns.pairplot(df.select_dtypes(exclude=(object)),hue='on_time')
plt.show()

<Figure size 1500x1500 with 0 Axes>

plt.figure(figsize=(15,15))
plt.subplot(3, 3, 1)
sns.scatterplot(data=df,x="Prior_purchases",y="Cost_of_the_Product",hue='on_time')
plt.title("Prior_purchases X Cost_of_the_Product")
plt.subplot(3, 3, 2)
sns.scatterplot(data=df,x="Prior_purchases",y="Weight_in_gms",hue='on_time')
plt.title("Prior_purchases X Weight_in_gms")
plt.subplot(3, 3, 3)
sns.scatterplot(data=df,x="Customer_care_calls",y="Prior_purchases",hue='on_time')
plt.title("Customer_care_calls X Prior_purchases ")
plt.subplot(3, 3, 4)
sns.scatterplot(data=df,x="Customer_care_calls",y="Weight_in_gms",hue='on_time')
plt.title("Customer_care_calls X Weight_in_gms")
plt.subplot(3, 3, 5)
sns.scatterplot(data=df,x="Customer_care_calls",y="Cost_of_the_Product",hue='on_time')
plt.title("Customer_care_calls X Cost_of_the_Product")
plt.subplot(3, 3, 6)
sns.scatterplot(data=df,x="Weight_in_gms",y="Discount_offered",hue='on_time')
plt.title("Weight_in_gms X Discount_offered")
#
plt.suptitle("Multivariable view")
plt.tight_layout()
plt.show()

df_vif = df.drop(columns='on_time')
vif_data = pd.DataFrame()
vif_data["Feature"] = df_vif.columns 
vif_data["VIF"] = [variance_inflation_factor(df_vif.values, i) for i in range(len(df_vif.columns))]
vif_data

plt.figure(figsize=(10,5))
plot = sns.barplot(data=vif_data,y='Feature',x='VIF',orient='h')
plot.axvline(x=5,color='red')
plot.axvline(x=vif_data['VIF'].mean(),color='green')
plt.title("VIF correlations")
plt.show()

df.insert(0,allow_duplicates=False,column='effective_cost',value = df['Cost_of_the_Product']* (1 - df['Discount_offered']/100))
df.insert(0,allow_duplicates=False,column='transport',value = pd.cut(df['Weight_in_gms'],right=False,bins=range(0,10000,2000),labels=range(0,4)))
df.insert(0,allow_duplicates=False,column='good_disc',value = pd.cut(df['Discount_offered'],right=False,bins=[0,10,100],labels=[0,1]))

mask = np.triu(np.ones_like(df.select_dtypes(exclude=object).corr(), dtype=np.bool_))
plt.figure(figsize=(15,15))
sns.heatmap(df.select_dtypes(exclude=object).corr(method='spearman'), mask=mask,fmt='.2f', square = True, annot=True, vmin=-1, vmax=1)
plt.show()

cols_model = ['transport', 
              'good_disc', 
              'effective_cost',
              'Warehouse_block', 
              'Mode_of_Shipment',
              #'Customer_care_calls',
              #'Cost_of_the_Product','Gendder,
              'Prior_purchases', 
              'Product_importance'
              ,'Discount_offered'
              #,'Weight_in_gms'
       ]

df_vif = df[cols_model]
vif_data = pd.DataFrame()
vif_data["Feature"] = df_vif.columns 
vif_data["VIF"] = [variance_inflation_factor(df_vif.values, i) for i in range(len(df_vif.columns))]
vif_data

plt.figure(figsize=(10,5))
plot = sns.barplot(data=vif_data,y='Feature',x='VIF',orient='h')
plot.axvline(x=5,color='red')
plot.axvline(x=vif_data['VIF'].mean(),color='green')
plt.title("VIF correlations")
plt.show()

col_scale = df.drop(columns=['on_time']).select_dtypes(exclude=object).columns
scale_index = np.where((df[col_scale].max() - df[col_scale].min()) > 10)
col_scale = col_scale[scale_index]

for col in col_scale:
    df[col] = StandardScaler().fit_transform(df[[col]])

joblib.dump(df,filename='./pkl/df_treat.pkl')

['df_treat.pkl']

df = joblib.load('./pkl/df_treat.pkl')

k = KFold(n_splits=5,random_state=42,shuffle=True)

cols_model = ['transport', 
              'good_disc', 
              'effective_cost',
              'Warehouse_block', 
              'Mode_of_Shipment',
              #'Customer_care_calls',
              #'Cost_of_the_Product','Gendder,
              'Prior_purchases', 
              'Product_importance'
              ,'Discount_offered'
              #,'Weight_in_gms'
       ]

ok = df[df['on_time']==1].index.to_list()
n_ok = df[df['on_time']==0].index.to_list()

lst_index = ok[0:len(n_ok)] + n_ok

X = df.loc[lst_index,cols_model].values
Y = df.loc[lst_index,'on_time'].values

joblib.dump((X,Y),"./pkl/X_Y.pkl")

['X_Y.pkl']

x_train, x_test, y_train, y_test = train_test_split(X,Y,test_size=0.10)
#cross_val = cross_val_score(X=X,y=Y,cv=k)

joblib.dump((x_train,x_test,y_train,y_test),"./pkl/train_test.pkl")

['train_test.pkl']

bayes = GaussianNB()
bayes.fit(X=x_train,y=y_train)
pred = bayes.predict(X=x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.7713963963963963
Classification
              precision    recall  f1-score   support

           0       0.70      0.91      0.79       420
           1       0.89      0.65      0.75       468

    accuracy                           0.77       888
   macro avg       0.79      0.78      0.77       888
weighted avg       0.80      0.77      0.77       888

mean abs error
0.2286036036036036
mean sqd error
0.2286036036036036
R²
0.08290598290598283

cm = ConfusionMatrix(bayes)
cm.fit(x_train,y_train)
cm.score(x_test,y_test)

0.7713963963963963

joblib.dump(bayes,filename='./pkl/bayes_model.pkl')

['bayes_model.pkl']

param = {'tol': [0.01, 0.001, 0.0001],
        'C': [1.0, 1.5, 2.0],
        'solver': ['lbfgs', 'sag', 'saga']}
test_grid = GridSearchCV(LogisticRegression(),param_grid=param)
test_grid.fit(X=x_train,y=y_train)
grid_pred = test_grid.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=grid_pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=grid_pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=grid_pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=grid_pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=grid_pred))
print("Best parameters")
print(test_grid.best_params_)

Accuracy
0.7804054054054054
Classification
              precision    recall  f1-score   support

           0       0.71      0.91      0.80       420
           1       0.89      0.67      0.76       468

    accuracy                           0.78       888
   macro avg       0.80      0.79      0.78       888
weighted avg       0.80      0.78      0.78       888

mean abs error
0.2195945945945946
mean sqd error
0.2195945945945946
R²
0.11904761904761896
Best parameters
{'C': 1.0, 'solver': 'saga', 'tol': 0.01}

linear_model = LogisticRegression(**test_grid.best_params_)

cross_val = cross_val_score(LogisticRegression(),X,Y,cv=k)
print(f"r²=\n\t{cross_val}\n mean r²:\n\t{np.mean(cross_val)}\n desv r²:\n\t{np.std(cross_val)}")

r²=
	[0.79774648 0.78028169 0.77903044 0.7886133  0.7897407 ]
 mean r²:
	0.787082522190641
 desv r²:
	0.006844216769937233

linear_model.fit(X=x_train,y=y_train)
pred = linear_model.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.7792792792792793
Classification
              precision    recall  f1-score   support

           0       0.70      0.92      0.80       420
           1       0.90      0.65      0.76       468

    accuracy                           0.78       888
   macro avg       0.80      0.79      0.78       888
weighted avg       0.81      0.78      0.78       888

mean abs error
0.22072072072072071
mean sqd error
0.22072072072072071
R²
0.11452991452991446

cm = ConfusionMatrix(linear_model)
cm.fit(x_train,y_train)
cm.score(x_test,y_test)

0.7792792792792793

joblib.dump(linear_model,filename='./pkl/linear_reg_model.pkl')

['linear_reg_model.pkl']

param = {
    'n_neighbors':[100,50,20,10],
    'weights':['uniform','distance'],
    'algorithm':['auto'],
    'p':[1,2]
    #metric='minkowski',
}
test_grid = GridSearchCV(KNeighborsClassifier(),param_grid=param)
test_grid.fit(x_train,y=y_train)
grid_pred = test_grid.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=grid_pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=grid_pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=grid_pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=grid_pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=grid_pred))
print("Best parameters")
print(test_grid.best_params_)

Accuracy
0.8288288288288288
Classification
              precision    recall  f1-score   support

           0       0.74      0.99      0.85       420
           1       0.99      0.68      0.81       468

    accuracy                           0.83       888
   macro avg       0.86      0.84      0.83       888
weighted avg       0.87      0.83      0.83       888

mean abs error
0.17117117117117117
mean sqd error
0.17117117117117117
R²
0.3133089133089132
Best parameters
{'algorithm': 'auto', 'n_neighbors': 20, 'p': 1, 'weights': 'uniform'}

knn_model = KNeighborsClassifier(**test_grid.best_params_)
cross_val = cross_val_score(knn_model,X,Y,cv=k)
print(f"r²=\n\t{cross_val}\n mean r²:\n\t{np.mean(cross_val)}\n desv r²:\n\t{np.std(cross_val)}")

r²=
	[0.84450704 0.82366197 0.81905299 0.83652762 0.83258174]
 mean r²:
	0.8312662718135192
 desv r²:
	0.009074358767660344

knn_model = KNeighborsClassifier(**test_grid.best_params_)

knn_model.fit(X=x_train,y=y_train)
pred = knn_model.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.8288288288288288
Classification
              precision    recall  f1-score   support

           0       0.74      0.99      0.85       420
           1       0.99      0.68      0.81       468

    accuracy                           0.83       888
   macro avg       0.86      0.84      0.83       888
weighted avg       0.87      0.83      0.83       888

mean abs error
0.17117117117117117
mean sqd error
0.17117117117117117
R²
0.3133089133089132

cm = ConfusionMatrix(knn_model)
cm.fit(x_train,y_train)
cm.score(x_test,y_test)

0.8288288288288288

joblib.dump(knn_model,filename='./pkl/knn_model.pkl')

['knn_model.pkl']

param = {
    'criterion':["gini", "entropy"],
    'splitter': ['best', 'random'],
    'min_samples_split':[2,5,15],
    'min_samples_leaf': [1,5,10],
    'random_state':[42]
}
test_grid = GridSearchCV(DecisionTreeClassifier(),param_grid=param)
test_grid.fit(X=x_train,y=y_train)
grid_pred = test_grid.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=grid_pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=grid_pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=grid_pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=grid_pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=grid_pred))
print("Best parameters")
print(test_grid.best_params_)

Accuracy
0.831081081081081
Classification
              precision    recall  f1-score   support

           0       0.75      0.95      0.84       420
           1       0.94      0.72      0.82       468

    accuracy                           0.83       888
   macro avg       0.85      0.84      0.83       888
weighted avg       0.85      0.83      0.83       888

mean abs error
0.16891891891891891
mean sqd error
0.16891891891891891
R²
0.3223443223443223
Best parameters
{'criterion': 'gini', 'min_samples_leaf': 10, 'min_samples_split': 2, 'random_state': 42, 'splitter': 'random'}

tree_model = DecisionTreeClassifier(**test_grid.best_params_)

cross_val = cross_val_score(DecisionTreeClassifier(),X,Y,cv=k)
print(f"r²=\n\t{cross_val}\n mean r²:\n\t{np.mean(cross_val)}\n desv r²:\n\t{np.std(cross_val)}")

r²=
	[0.77746479 0.77239437 0.75930101 0.76662909 0.76493799]
 mean r²:
	0.7681454499261635
 desv r²:
	0.006256429269677034

tree_model.fit(X=x_train,y=y_train)
pred = tree_model.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.831081081081081
Classification

              precision    recall  f1-score   support

           0       0.75      0.95      0.84       420
           1       0.94      0.72      0.82       468

    accuracy                           0.83       888
   macro avg       0.85      0.84      0.83       888
weighted avg       0.85      0.83      0.83       888

mean abs error
0.16891891891891891
mean sqd error
0.16891891891891891
R²
0.3223443223443223

cm = ConfusionMatrix(tree_model)
cm.fit(x_train,y_train)
cm.score(x_test,y_test)

0.831081081081081

joblib.dump(tree_model,filename='./pkl/dec_tree_model.pkl')

['dec_tree_model.pkl']

param = {
    'criterion':["gini", "entropy"],
    'n_estimators':[50,100,200],
    'min_samples_split':[2,5,10],
    'min_samples_leaf': [1,5,10],
    'random_state':[42]
}
test_grid = GridSearchCV(RandomForestClassifier(),param_grid=param)
test_grid.fit(X=x_train,y=y_train)
grid_pred = test_grid.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=grid_pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=grid_pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=grid_pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=grid_pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=grid_pred))
print("Best parameters")
print(test_grid.best_params_)

Accuracy
0.8412162162162162
Classification
              precision    recall  f1-score   support

           0       0.75      0.99      0.85       420
           1       0.99      0.71      0.82       468

    accuracy                           0.84       888
   macro avg       0.87      0.85      0.84       888
weighted avg       0.88      0.84      0.84       888

mean abs error
0.15878378378378377
mean sqd error
0.15878378378378377
R²
0.36300366300366294
Best parameters
{'criterion': 'entropy', 'min_samples_leaf': 5, 'min_samples_split': 2, 'n_estimators': 200, 'random_state': 42}

rf_model = RandomForestClassifier(**test_grid.best_params_)

cross_val = cross_val_score(rf_model,X,Y,cv=k)
print(f"r²=\n\t{cross_val}\n mean r²:\n\t{np.mean(cross_val)}\n desv r²:\n\t{np.std(cross_val)}")

r²=
	[0.85915493 0.84394366 0.82919955 0.85005637 0.84949267]
 mean r²:
	0.8463694364609301
 desv r²:
	0.009874272907078164

rf_model.fit(X=x_train,y=y_train)
pred = rf_model.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.8412162162162162
Classification
              precision    recall  f1-score   support

           0       0.75      0.99      0.85       420
           1       0.99      0.71      0.82       468

    accuracy                           0.84       888
   macro avg       0.87      0.85      0.84       888
weighted avg       0.88      0.84      0.84       888

mean abs error
0.15878378378378377
mean sqd error
0.15878378378378377
R²
0.36300366300366294

cm = ConfusionMatrix(rf_model)
cm.fit(x_train,y_train)
cm.score(x_test,y_test)

0.8412162162162162

joblib.dump(rf_model,filename='./pkl/rf_model.pkl')

['rf_model.pkl']

param = {'tol': [0.1, 0.01, 0.001],
         'C': [1.0,2.0],
         'kernel': ['rbf', 'linear', 'poly', 'sigmoid'],
         'random_state':[42]
         }
test_grid = GridSearchCV(SVC(),param_grid=param)
test_grid.fit(X=x_train,y=y_train)
grid_pred = test_grid.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=grid_pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=grid_pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=grid_pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=grid_pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=grid_pred))
print("Best parameters")
print(test_grid.best_params_)

Accuracy
0.8412162162162162
Classification
              precision    recall  f1-score   support

           0       0.75      0.99      0.85       420
           1       0.99      0.71      0.82       468

    accuracy                           0.84       888
   macro avg       0.87      0.85      0.84       888
weighted avg       0.88      0.84      0.84       888

mean abs error
0.15878378378378377
mean sqd error
0.15878378378378377
R²
0.36300366300366294
Best parameters
{'C': 2.0, 'kernel': 'rbf', 'random_state': 42, 'tol': 0.01}

svc_model = SVC(**test_grid.best_params_)

cross_val = cross_val_score(svc_model,X,Y,cv=k)
print(f"r²=\n\t{cross_val}\n mean r²:\n\t{np.mean(cross_val)}\n desv r²:\n\t{np.std(cross_val)}")

r²=
	[0.85464789 0.8428169  0.82976325 0.85005637 0.85005637]
 mean r²:
	0.845468155040729
 desv r²:
	0.008718304874540717

svc_model.fit(X=x_train,y=y_train)
pred = svc_model.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.8603603603603603
Classification
              precision    recall  f1-score   support

           0       0.78      0.99      0.87       436
           1       0.99      0.74      0.84       452

    accuracy                           0.86       888
   macro avg       0.88      0.86      0.86       888
weighted avg       0.89      0.86      0.86       888

mean abs error
0.13963963963963963
mean sqd error
0.13963963963963963
R²
0.4412600470893886

cm = ConfusionMatrix(svc_model)
cm.fit(x_train,y_train)
cm.score(x_test,y_test)

0.8603603603603603

joblib.dump(svc_model,filename='./pkl/svc_model.pkl')

['svc_model.pkl']

param = {
    'hidden_layer_sizes':[(100),(10,10),(8,32,8)],
    'activation': ['relu', 'logistic'],
    'solver': ['adam', 'sgd'],
    'max_iter':[200],
    'verbose':[False],
    'tol':[0.0000010],
    'random_state': [42],
    'learning_rate':["constant", "invscaling", "adaptive"]
    ,'batch_size': ['auto']
}
test_grid = GridSearchCV(MLPClassifier(),param_grid=param,n_jobs=-1)
test_grid.fit(X=x_train,y=y_train)
grid_pred = test_grid.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=grid_pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=grid_pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=grid_pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=grid_pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=grid_pred))
print("Best parameters")
print(test_grid.best_params_)

Accuracy
0.8637387387387387
Classification
              precision    recall  f1-score   support

           0       0.79      0.99      0.88       436
           1       0.99      0.74      0.85       452

    accuracy                           0.86       888
   macro avg       0.89      0.87      0.86       888
weighted avg       0.89      0.86      0.86       888

mean abs error
0.13626126126126126
mean sqd error
0.13626126126126126
R²
0.4547779491759356
Best parameters
{'activation': 'relu', 'batch_size': 'auto', 'hidden_layer_sizes': (10, 10), 'learning_rate': 'constant', 'random_state': 42, 'solver': 'adam', 'tol': 1e-06, 'verbose': False}

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

nn_model = MLPClassifier(**test_grid.best_params_)

cross_val = cross_val_score(nn_model,X,Y,cv=k)
print(f"r²=\n\t{cross_val}\n mean r²:\n\t{np.mean(cross_val)}\n desv r²:\n\t{np.std(cross_val)}")

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

r²=
	[0.85183099 0.84225352 0.82863585 0.85005637 0.84723788]
 mean r²:
	0.8440029217015736
 desv r²:
	0.008337805947826288

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

nn_model.fit(X=x_train,y=y_train)
pred = nn_model.predict(x_test)
print("Accuracy")
print(accuracy_score(y_true=y_test,y_pred=pred))
print("Classification")
print(classification_report(y_true=y_test,y_pred=pred))
print("mean abs error")
print(mean_absolute_error(y_true=y_test,y_pred=pred))
print("mean sqd error")
print(mean_squared_error(y_true=y_test,y_pred=pred))
print("R²")
print(r2_score(y_true=y_test,y_pred=pred))

Accuracy
0.8637387387387387
Classification
              precision    recall  f1-score   support

           0       0.79      0.99      0.88       436
           1       0.99      0.74      0.85       452

    accuracy                           0.86       888
   macro avg       0.89      0.87      0.86       888
weighted avg       0.89      0.86      0.86       888

mean abs error
0.13626126126126126
mean sqd error
0.13626126126126126
R²
0.4547779491759356

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

cm = ConfusionMatrix(nn_model)
cm.fit(x_train,y_train)
'cm.score(x_test,y_test)

0.8637387387387387

joblib.dump(nn_model,filename='./pkl/nn_model.pkl')

['nn_model.pkl']

bayes_model = joblib.load('./pkl/bayes_model.pkl')
tree_model = joblib.load('./pkl/dec_tree_model.pkl')
knn_model = joblib.load('./pkl/knn_model.pkl')
rf_model = joblib.load('./pkl/rf_model.pkl')
linear_model = joblib.load('./pkl/linear_reg_model.pkl')
svc_model = joblib.load('./pkl/svc_model.pkl')
nn_model = joblib.load('./pkl/nn_model.pkl')

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\base.py:380: InconsistentVersionWarning: Trying to unpickle estimator GaussianNB from version 1.7.0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\base.py:380: InconsistentVersionWarning: Trying to unpickle estimator DecisionTreeClassifier from version 1.7.0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\base.py:380: InconsistentVersionWarning: Trying to unpickle estimator KNeighborsClassifier from version 1.7.0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\base.py:380: InconsistentVersionWarning: Trying to unpickle estimator RandomForestClassifier from version 1.7.0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\base.py:380: InconsistentVersionWarning: Trying to unpickle estimator LogisticRegression from version 1.7.0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\base.py:380: InconsistentVersionWarning: Trying to unpickle estimator SVC from version 1.7.0 when using version 1.6.1. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/model_persistence.html#security-maintainability-limitations
  warnings.warn(

X, Y = joblib.load('./pkl/X_Y.pkl')
x_train, x_test, y_train, y_test = joblib.load('./pkl/train_test.pkl')

tree_results = []
random_forest_results = []
knn_results = []
linear_result = []
svm_result = []
nn_results = []
bayes_result = []

for i in range(30):
  print(f'Teste: {i}')
  # Tree model
  kfold = KFold(n_splits=5, shuffle=True, random_state=i)
  # bayes
  print(f'-- Bayes')
  scores = cross_val_score(bayes_model, X, Y, cv = kfold)
  bayes_result.append(scores.mean())
  # decision tree
  print(f'-- Trees')
  scores = cross_val_score(tree_model, X, Y, cv = kfold)
  tree_results.append(scores.mean())
  # random forest
  print(f'-- Random Rorest')
  scores = cross_val_score(rf_model,  X, Y, cv = kfold)
  random_forest_results.append(scores.mean())
  #KNN
  print(f'-- KNN')
  scores = cross_val_score(knn_model, X, Y, cv = kfold)
  knn_results.append(scores.mean())
  #Linear regression
  print(f'-- Linear regression')
  scores = cross_val_score(linear_model, X, Y, cv = kfold)
  linear_result.append(scores.mean())
  #SVC
  print(f'-- svc')
  scores = cross_val_score(svc_model, X, Y, cv = kfold)
  svm_result.append(scores.mean())
  # Neural Network
  print(f'-- Neural Network')
  scores = cross_val_score(nn_model, X, Y, cv = kfold)
  nn_results.append(scores.mean())

Teste: 0
-- Bayes
-- Trees
-- Random Rorest
-- KNN
-- Linear regression
-- svc
-- Neural Network

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

Teste: 1
-- Bayes
-- Trees
-- Random Rorest
-- KNN
-- Linear regression
-- svc
-- Neural Network

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

Teste: 2
-- Bayes
-- Trees
-- Random Rorest
-- KNN
-- Linear regression
-- svc
-- Neural Network

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

Teste: 3
-- Bayes
-- Trees
-- Random Rorest
-- KNN
-- Linear regression
-- svc
-- Neural Network

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(
C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

Teste: 4
-- Bayes
-- Trees
-- Random Rorest
-- KNN
-- Linear regression
-- svc
-- Neural Network

C:\Users\ThalesFerreira\AppData\Roaming\Python\Python312\site-packages\sklearn\neural_network\_multilayer_perceptron.py:691: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (200) reached and the optimization hasn't converged yet.
  warnings.warn(

df_resultados = pd.DataFrame(
    {"bayes":bayes_result,'linear':linear_result,'knn':knn_results,
     "tree":tree_results,'random_forest':random_forest_results,
     'svc':svm_result,'neural_net':nn_results}
)

joblib.dump(df_resultados,filename='df_results.pkl')

['df_results.pkl']

df_resultados.mean()

bayes            0.766761
linear           0.788721
knn              0.831117
tree             0.833934
random_forest    0.846660
svc              0.845638
neural_net       0.844210
dtype: float64

df_resultados.std()

bayes            0.000170
linear           0.000809
knn              0.000781
tree             0.002340
random_forest    0.000450
svc              0.000205
neural_net       0.000485
dtype: float64

for col in df_resultados.columns:
    print(shapiro(df_resultados[col]))

ShapiroResult(statistic=np.float64(0.9428308320517239), pvalue=np.float64(0.10842600092552829))
ShapiroResult(statistic=np.float64(0.958842646295605), pvalue=np.float64(0.28928784404866514))
ShapiroResult(statistic=np.float64(0.9817636637271815), pvalue=np.float64(0.8702608042442374))
ShapiroResult(statistic=np.float64(0.9693575036013388), pvalue=np.float64(0.5217474528397927))
ShapiroResult(statistic=np.float64(0.9784638643703881), pvalue=np.float64(0.7834009057822708))
ShapiroResult(statistic=np.float64(0.9427677444389647), pvalue=np.float64(0.10800219793792448))
ShapiroResult(statistic=np.float64(0.9704187344533626), pvalue=np.float64(0.5506958651692518))

for i,col in enumerate(df_resultados.columns):
    sns.displot(df_resultados[col].values, kind = 'kde')
    plt.xlabel('Accuracy')
    plt.title(col)

_, p = f_oneway(tree_results,random_forest_results,knn_results,linear_result,svm_result,nn_results,bayes_result)
p

np.float64(5.090282508519619e-295)

if p <= 0.05:
  print('Data are differents')
else:
  print('The results are equals')

Data are differents

df_result_melt = pd.melt(df_resultados,var_name='model',value_name='accuracy')

model_compare = MultiComparison(df_result_melt['accuracy'], df_result_melt['model'])

statistic_test = model_compare.tukeyhsd()
print(statistic_test)

       Multiple Comparison of Means - Tukey HSD, FWER=0.05        
==================================================================
    group1        group2    meandiff p-adj   lower   upper  reject
------------------------------------------------------------------
        bayes           knn   0.0644    0.0  0.0636  0.0651   True
        bayes        linear    0.022    0.0  0.0212  0.0227   True
        bayes    neural_net   0.0774    0.0  0.0767  0.0782   True
        bayes random_forest   0.0799    0.0  0.0791  0.0807   True
        bayes           svc   0.0789    0.0  0.0781  0.0797   True
        bayes          tree   0.0672    0.0  0.0664   0.068   True
          knn        linear  -0.0424    0.0 -0.0432 -0.0416   True
          knn    neural_net   0.0131    0.0  0.0123  0.0139   True
          knn random_forest   0.0155    0.0  0.0148  0.0163   True
          knn           svc   0.0145    0.0  0.0137  0.0153   True
          knn          tree   0.0028    0.0   0.002  0.0036   True
       linear    neural_net   0.0555    0.0  0.0547  0.0563   True
       linear random_forest   0.0579    0.0  0.0572  0.0587   True
       linear           svc   0.0569    0.0  0.0561  0.0577   True
       linear          tree   0.0452    0.0  0.0444   0.046   True
   neural_net random_forest   0.0024    0.0  0.0017  0.0032   True
   neural_net           svc   0.0014    0.0  0.0006  0.0022   True
   neural_net          tree  -0.0103    0.0 -0.0111 -0.0095   True
random_forest           svc   -0.001 0.0026 -0.0018 -0.0002   True
random_forest          tree  -0.0127    0.0 -0.0135 -0.0119   True
          svc          tree  -0.0117    0.0 -0.0125 -0.0109   True
------------------------------------------------------------------

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
ID	10999.0	NaN	NaN	NaN	5500.0	3175.28214	1.0	2750.5	5500.0	8249.5	10999.0
Warehouse_block	10999	5	F	3666	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Mode_of_Shipment	10999	3	Ship	7462	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Customer_care_calls	10999.0	NaN	NaN	NaN	4.054459	1.14149	2.0	3.0	4.0	5.0	7.0
Customer_rating	10999.0	NaN	NaN	NaN	2.990545	1.413603	1.0	2.0	3.0	4.0	5.0
Cost_of_the_Product	10999.0	NaN	NaN	NaN	210.196836	48.063272	96.0	169.0	214.0	251.0	310.0
Prior_purchases	10999.0	NaN	NaN	NaN	3.567597	1.52286	2.0	3.0	3.0	4.0	10.0
Product_importance	10999	3	low	5297	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Gender	10999	2	F	5545	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Discount_offered	10999.0	NaN	NaN	NaN	13.373216	16.205527	1.0	4.0	7.0	10.0	65.0
Weight_in_gms	10999.0	NaN	NaN	NaN	3634.016729	1635.377251	1001.0	1839.5	4149.0	5050.0	7846.0
Reached.on.Time_Y.N	10999.0	NaN	NaN	NaN	0.596691	0.490584	0.0	0.0	1.0	1.0	1.0

	ID	Warehouse_block	Mode_of_Shipment	Customer_care_calls	Customer_rating	Cost_of_the_Product	Prior_purchases	Product_importance	Gender	Discount_offered	Weight_in_gms	Reached.on.Time_Y.N
1697	1698	F	Ship	3	2	263	3	high	F	44	3802	1
2536	2537	C	Ship	4	4	205	3	low	F	58	2522	1
3513	3514	B	Ship	3	2	262	4	low	M	1	4745	1
6586	6587	C	Flight	6	4	299	4	medium	M	9	1036	1
2215	2216	F	Ship	3	4	260	4	high	M	17	2666	1

	Feature	VIF
0	Warehouse_block	3.350684
1	Mode_of_Shipment	4.733955
2	Customer_care_calls	14.182232
3	Customer_rating	5.173562
4	Cost_of_the_Product	18.247517
5	Prior_purchases	6.399117
6	Product_importance	5.211311
7	Gender	1.957789
8	Discount_offered	1.735545
9	Weight_in_gms	5.324983

	Feature	VIF
0	transport	3.098094
1	good_disc	3.203416
2	effective_cost	1.954329
3	Warehouse_block	3.101051
4	Mode_of_Shipment	4.176865
5	Prior_purchases	4.642797
6	Product_importance	4.508913
7	Discount_offered	3.303627

Import Libs¶

System¶

Data¶

Import libs to analyse data¶

Models¶

Import libs divinding data and analyse models results¶

Statistical validation¶

Import libs models¶

Import Data¶

Explore Data pt 1¶

Treat pt 1¶

Data Visualisation¶

Count on_time¶

View data frequency¶

on_time relation visualization¶

Numeric Columns¶

Category columns¶

Outliers¶

Treat pt 2¶

Multicolumns verification¶

Spearman Relation¶

Multivariable view¶

VIF¶

Treat pt 3¶

New features¶

Spearman Relation¶

Columns¶

VIF after features¶

StandardScaler¶

Base to train¶

Save df treated¶

load data¶

Split base¶

Model training¶

Bayes¶

Linear regression¶

KNN¶

Decision Tree¶

Random Forest¶

SVM¶

Neural Network¶

Testes¶

Load Models¶

Cross validations¶

Statistical Test¶

Shapiro¶

normality¶

anova one_way¶

turkey¶